Introduction: The multidisciplinary nature of ironman (IM) triathlon races makes the analysis of performance more challenging. However, an analysis of race patterns (i.e., the real contribution of each discipline to a good ranking overall) is needed for athletes and coaches to plan training and race strategy. To our knowledge, only the timed contribution of each discipline to overall time is classically used to describe race patterns in ironman triathlon races but this data may not be of relevance as it strongly depends on the event and overall time. Here we proposed to analyze IM triathlon race results using a multidimensional descriptive approach and to provide a simple graphical representation of an athlete`s race structure as well as its correlation with overall performance. This new method allows the identification of successful (and unsuccessful) race patterns. Methods: We used race results of IM Hawaii 2012 to compute a principal component analysis (PCA) by group division (Pro and age groupers) and gender. All variables related to swim time, bike time, run time and transition time were taken into account and names of the competitors were replaced by their final ranking. All results and graphs were obtained with SAS 9.3 (SAS Institute Inc., Cary, NC, USA). Original data was taken from « slowtwitch enhanced results » (available online : http://www.slowtwitch.com/enhancedresults/). Results: PCA of IM Hawaii triathletes revealed a strong correlation of bike-related variables with the 1st axis (rc1=0.45) while run-related variables and swim-related variables were respectively correlated and negatively correlated with axis 2 (rc2=0.19). When athletes were projected within those axis, a clear ranking-dependent pattern appeared : the best triathletes were characterized by a balanced performance in all three disciplines, with a high prevalence of top running performance. After a certain rank, race patterns were characterized by a slower time in one of the three disciplines, more than a moderate performance in all three disciplines. Finally, combining a slow bike split with another slow split in one of the 2 other disciplines (especially running) was the main reason for a low final ranking. Eventually, category-specific patterns were found such as the absence of data structure linking transition times and final ranking in the professional category, whereas this structure appeared in several age groups. Discussion: PCA is a descriptive method that is not based on probabilistic assumptions but on a geometric model. Further investigation can be done using the race patterns described here using inferential statistics. However, the graphical data generated by PCA can be used to point out the existence of classical race patterns associated with performance as well as to describe an athlete`s performance after the race. We believe that such data may be of relevance for triathletes and coaches to adapt their training programs.
© Copyright 2014 19th Annual Congress of the European College of Sport Science (ECSS), Amsterdam, 2. - 5. July 2014. Veröffentlicht von VU University Amsterdam. Alle Rechte vorbehalten.